Evidence before execution
RunOnProof helps software agents decide whether a company-related action has enough evidence to continue. It does not return a vague trust score. It identifies the legal business, checks the evidence available for a named action, applies a declared policy, and returns a narrow result with its reasons and limits attached.
Imagine an agent that wants to add a supplier, accept an invoice, change bank details, or release a payment. Finding a company record is only the beginning. The agent still needs to know whether it found the right legal entity, whether the source is current, which facts were actually verified, and whether those facts satisfy the rule for this specific action. RunOnProof packages those questions into decisions that both machines and people can inspect.
The simple version
RunOnProof follows five steps:
- Understand the action. Adding a supplier is different from paying an invoice.
- Find the legal company. A trading name is not always the registered entity.
- Collect scoped evidence. Each source supports only certain claims.
- Apply the named rule. Evidence is compared with the policy for that action.
- Return a decision. The result is PROCEED, REVIEW, or STOP, with evidence and limitations.
The claim is never meant to be larger than the evidence. If a source is unavailable, stale, ambiguous, or outside its declared coverage, that uncertainty remains visible.
Who this documentation is for
You can read these pages without being a developer. Product and operations teams can use them to understand what RunOnProof does and does not approve. Developers can use the integration pages to connect an API or agent platform. Risk and compliance teams can use the coverage, decision-state, and enterprise pages to review boundaries before production use.
Start with the question you need answered
| Your question | Where to begin |
|---|---|
| What does RunOnProof sell? | Catalog |
| Can I use it in a specific country? | Coverage and the country page |
| What do PROCEED, REVIEW, and STOP mean? | Decision states |
| How do I test safely? | Get started and Synthetic demo |
| How do I connect an agent? | Integrations |
| How do I operate it inside a company? | Enterprise operations |
What RunOnProof is not
RunOnProof is not a universal company score, a credit bureau, a bank, an accounting system, or a promise that fraud is impossible. It does not turn one successful check into permission for every future action. It does not silently replace a customer's policy, and an AI model is not allowed to invent missing evidence.
A decision is useful because its scope is narrow. “This legal entity was active in an official registry at this observation time” is different from “this company is safe.” “This payee matched under the available binding evidence” is different from “every payment to this account is authorized forever.”
The rule that governs every page
Documentation explains the model, but live operational truth comes from the production catalog, coverage manifest, source-health information, and release identity. If a historical page and runtime truth disagree, runtime truth wins. Read Status before relying on a capability in production.